Summary The preoperative concentrations of IgG were lower (P<0.002) and the concentrations of C4 and C1-INH higher (P<0.01 and P<0.001) in 29 patients with recurrence after potentially curative resection of gastric carcinoma, than in 31 patients alive and disease-free 5 years after surgery. These differences between the two groups of patients were consistent within each of six groups of disease extent. In each of the two groups of patients, the preoperative concentrations of IgG, C4 and Cl-INH had no significant variation with the extent of disease (P>0.05 or greater). Of our variables, Cl-INH was the most potent prognosticator and discriminated between patients with and without recurrence with 80% accuracy. Furthermore, the predictive prognostic value of Cl-INH at the time of surgery was superior to the prognostic value of the extent of disease (F values 27.00 and 12.69). Apparently, the preoperative Cl-INH concentration is an essential and independent prognostic parameter of gastric carcinoma. We assume that Cl-INH reflects an additional prognostic feature appropriate to the tumour or the host. Our finding that the interval between surgery and death from recurrence had an inverse relation to the preoperative Cl-INH concentration also supports this assumption.We have previously described the preoperative serum concentrations of immunoglobulins (Ig) and some complement components (C), and the erythrocyte sedimentation rate (ESR) in relation to the extent of disease and prognosis in patients with gastric carcinoma (Janssen et al., 1983(Janssen et al., , 1985(Janssen et al., , 1987a. The levels of IgG decreased with advancing disease among those resected for cure, whereas Cl-esterase inhibitor (Cl-INH), C4 and ESR increased. Patients with recurrence had lower preoperative concentrations of lgG and higher concentrations of Cl-INH than those alive and disease-free 21 years after surgery.The series of patients is now extended and the patients have been followed for 5 years, which permits the retrieval of further results. We have now searched for the exact predictive value of our variables as to recurrence or no-recurrence.